Regulators publish documents, but financial institutions are judged on how well they manage the individual obligations buried inside them. That mismatch is where much of the cost and risk in regulatory change management sits, and it is the problem AscentAI, a regulatory intelligence platform, has built its approach around.
As part of its research for the AI in Risk and Compliance 2026 report, Parker & Lawrence Research spoke with several technology providers developing tools to help firms strengthen risk identification, assessment and oversight.
As part of The Demo Room series, the team spoke to AscentAI, a RegTech helping to automate the full regulatory lifecycle.
A single consultation, speech, rule or enforcement notice can contain a host of distinct requirements. Some introduce something new, others amend an existing position, and many have no practical bearing on a given firm, it said. A document-level alert answers none of these questions. While AI can speed up reading and comparison, regulatory interpretation carries a long tail of ambiguity. A summary can be accurate and still operationally useless if it fails to pinpoint the affected requirement, link it to the firm’s regulatory framework or send it to the people responsible for acting on it.
The industry appears acutely aware of this tension. According to 2026 research, 36% of institutions have deployed AI at scale in compliance management, yet 57% rate its AI-related risk as high or greater. Firms recognise the upside, but they also know that a missed obligation or a poorly evidenced interpretation can ripple through policies and controls, it said.
Conventional horizon-scanning tools gather publications and notify users, but the heavy lifting starts once the alert lands. Specialists must locate relevant passages, distinguish requirements from commentary, compare new wording against the existing rulebook and judge whether products, entities or processes are affected.
The findings then need mapping to owners, policies and controls, frequently across disconnected systems. At scale, this becomes repetitive and hard to evidence. Analysts may read the same change differently, ownership records can be patchy, and conclusions logged in workflow tools can drift away from the text that supports them. Global firms face the added burden of inconsistent formats, terminology and change patterns across regulators.
Placing a general-purpose language model at the front of this process may make the first read quicker. It does not, however, deliver a controlled regulatory data model or a dependable corporate regulatory map, meaning the firm’s record of which obligations apply to which business units and owners. Nor does it ensure that a generated answer will hold up when challenged by auditors or supervisors.
AscentAI’s response is to treat the obligation as the core unit of change. The platform ingests millions of lines of regulatory text spanning the US, UK, Europe and other jurisdictions, using proprietary machine-learning and natural-language-processing models to parse and classify it. Requirements are pulled out as individual obligation objects, each keeping its link to the original rule.
From there, the system compares new and existing obligations to show precisely how a requirement has shifted. Affected obligations are mapped to the customer’s corporate regulatory map, helping teams identify the business units, owners and requirements that may need attention.
The company deliberately assigns different technologies to different jobs. Machine-learning models handle bounded, high-volume tasks where control and consistency are paramount, with subject-matter experts adding a further layer of validation. Language models are used selectively for summarisation, comparison and natural-language retrieval, with outputs grounded in AscentAI’s own structured regulatory content.
Delivery is designed to fit existing operating models. Users can work within AscentAI, push intelligence into a GRC platform or access it through their own AI environment via an MCP server. Source-linked obligations, change records, ownership and review history create a traceable chain from publication through to assessment and downstream action, without forcing customers to replace their current GRC workflows or interfaces.
Analysts at Parker & Lawrence see AscentAI’s choice of unit as its clearest strength. Because the obligation is the level at which compliance teams ultimately need to understand and manage change, structuring regulatory material this way offers a firmer foundation for comparison, impact assessment and auditability than a stream of document summaries. By keeping structured regulatory data at the heart of the process, the design aims to give firms speed without sacrificing control.
For more insights into AI in risk and compliance, read the full report here.
Copyright © 2026 RegTech Analyst
Copyright © 2026 RegTech Analyst
